Lifting The Fog Of Code
The We Are Developers conference took place in the beautiful Hoffburg Kongress venue, a part of the imperial Hoffburg palace. It was a massive event, with thousands of people in the main theater. It lasted two days, talks were spread over four tracks, plus hands-on technical workshops. Presentations were focused on today's industry trends as well as what's coming in the near future.
AI & ML - The Plateau of Productivity Within Reach
AI is maturing and entering the mainstream. Good industry practices around AI/ML products are taking shape. Talks from industry insiders showcased the actual products, both in-house and publicly available, academic researchers showcased their recent work and publications.
The conference kicked-off with an amazing presentation by Tanmay Bakshi, a 16 year old wunderkind, IBM Cloud Advisor and AI expert. His talk was a show of force of technical savvy in signal processing, ML&AI and computer engineering which climaxed with a demo implementation of HeartID (similar to fingerprint ID on phones, but using electrocardiogram instead of a fingerprint). By using a 20 second long ECG of a person, he was able to uniquely identify them. The small ECG device he used for the demo communicates with the phone through, of all things, sound.
How computer vision is being utilized for tagging and sorting images at scale was at the heart of Noa Barbiro’s talk, representative of Booking.com. They utilize AI to get a deeper understanding of user uploaded photos. Those beautiful hotel cover photos on their website with captions like "great view!" or "good vibe!", are automatically selected by their AI models.
Product lifecycle played a prominent role in talks. Manfred Kügel focused his whole talk "How to make money with data analytics in industrial use cases" solely on that issue. How to move forward from a promising Proof of Concept model into something actually productive. How to manage a model repository, choose the right model for the right job, monitor and evaluate model performance in production, whether to retrain or retire a production model. He also touched on how to organize model training, assemble interdisciplinary AI teams made of data scientists, domain experts and software engineers.
Looking to the future, research brings incremental improvements to AI and makes it more approachable for the dev community. Hardware capability increases with every new generation, making the more elaborate and capable neural networks possible in an everyday setting. Self-training with generative adversarial networks is also taking off, but still finding its place. Pitfalls remain the same for now; still very narrow applications without a "one size fits all" solution. Accuracy and speed are improving, but still not 100% there. Training still requires expensive hardware and human resources/intervention.
Cloud Computing & Internet of Things (IoT) - "There and back again"
With cloud computing in full swing, providers are looking ahead of the curve: Edge computing, IoT, simplifying the operational security, 5G network applications.
"We spent all this time moving our IT infrastructure to the cloud, and now we’re moving it back???" - said Josef Hammer in his "Are you on the Edge? Or still in the cloud?" talk. Industry loves the flexibility of “Serverless”. Cloud providers are now pushing the Serverless capability to topological edges of networks to offer less latency and more bandwidth and privacy. Right now, processing done on personal devices is on the rise and more and more computing will be done in our close proximity, e.g. on a home server, gateway or a mobile provider’s base station.
This is especially important once IoT & AI applications start to go mainstream. Autonomous driving/flying requires very low latency and also privacy with so much data collection going on. Privacy concerns need to be addressed before “house robots” (vacuum cleaners, home assistants, smart security cameras etc.) can be fully adopted. Cloud providers are starting to offer everything from specially designed IoT chips to private “on-premises” servers. Mobile networks providers are also offering, through their APIs, “lambda” computing done on cell phone base stations in close proximity. Consumers are getting the best of both worlds with this approach: the convenience of cloud software and the best of Edge.
The speakers left us wanting more, they didn’t reveal how the portal app integrates widgets. They praised the ubiquitous scalability of their solution:
- Development perspective - they can parallelize the work with every team working independently on their own widget
- Investment perspective - The cost of adding a new feature stays constant no matter the size of application in direct contrast to monolith architecture
Operational security is getting complex: microservices use a wide variety of software packages and their versions -- personal data is flowing from devices and sensors, stored in the cloud. On the other hand, the sophistication of cyber attacks is increasing. We heard several talks from academia and industry leaders about leveraging AI for detecting attacks, code scanning and testing. Christopher Feussner, from Security Engineer Palo Alto Networks, showcased a product called Prisma Cloud and talked about how to simplify container security with it.
On the surface, this conference looked oversaturated by PR, recruiting and networking efforts, but we quickly changed our minds. We met some cool people, saw the trends, and got inspired. Events like this are lifting the “Fog of Code”. As their motto says: “you can’t download the experience.” We sure are glad we didn’t.